```json { "table_name": "column_taxonomy", "table_type_tags": [ {"tag": "dimension", "confidence": 0.95} ], "primary_type": "dimension", "insufficient_evidence": false, "entity": { "columns": ["jobs_id", "columns_catalog_id"], "unique_per_row": true }, "time_axis": null, "reasoning": "The composite entity (jobs_id, columns_catalog_id) is unique per row: jobs_id has RPUV 43.51 and columns_catalog_id has Uniqueness 0.968, and their product (116 × 4887 ≈ 567K) far exceeds the 5,047 row count, confirming uniqueness. The four provenance columns (_ingestion_timestamp, _batch_id, _source_file, _source_system) are synthetic load-time metadata and carry no evidence. No usable time axis exists — the table describes taxonomic classifications (family, subtype, measurement_level, aggregation_profile) of columns within jobs, with each row being unique descriptive information about one (job, column) pair." } ```